A triglycyl peptide linker (CX) was designed for use in antibody–drug conjugates (ADC), aiming to provide efficient release and lysosomal efflux of cytotoxic catabolites within targeted cancer cells. ADCs comprising anti-epithelial cell adhesion molecule (anti-EpCAM) and anti-EGFR antibodies with maytansinoid payloads were prepared using CX or a noncleavable SMCC linker (CX and SMCC ADCs). The in vitro cytotoxic activities of CX and SMCC ADCs were similar for several cancer cell lines; however, the CX ADC was more active (5–100-fold lower IC50) than the SMCC ADC in other cell lines, including a multidrug-resistant line. Both CX and SMCC ADCs showed comparable MTDs and pharmacokinetics in CD-1 mice. In Calu-3 tumor xenografts, antitumor efficacy was observed with the anti-EpCAM CX ADC at a 5-fold lower dose than the corresponding SMCC ADC in vivo. Similarly, the anti-EGFR CX ADC showed improved antitumor activity over the respective SMCC conjugate in HSC-2 and H1975 tumor models; however, both exhibited similar activity against FaDu xenografts. Mechanistically, in contrast with the charged lysine-linked catabolite of SMCC ADC, a significant fraction of the carboxylic acid catabolite of CX ADC could be uncharged in the acidic lysosomes, and thus diffuse out readily into the cytosol. Upon release from tumor cells, CX catabolites are charged at extracellular pH and do not penetrate and kill neighboring cells, similar to the SMCC catabolite. Overall, these data suggest that CX represents a promising linker option for the development of ADCs with improved therapeutic properties. Mol Cancer Ther; 15(6); 1311–20. ©2016 AACR.
Antibody–drug conjugates (ADC) are a broad category of therapeutic agents, each targeting a specific antigen expressed on the surface of cancer cells (1, 2). An ADC consists of three components: a monoclonal antibody that binds selectively to an antigen, a cytotoxic payload that inhibits essential cellular functions, and a linker that connects the two while ideally facilitating the intracellular release of the payload from the conjugate. Two ADCs are currently approved by the FDA: (1) brentuximab vedotin (Adcetris), consisting of an anti-CD30 antibody linked to a monomethyl auristatin (MMAE) payload, and (2) ado-trastuzumab emtansine (Kadcyla), consisting of the anti-HER2 antibody trastuzumab linked to a maytansinoid (DM1) payload (3, 4). In addition, about 50 ADCs targeting numerous antigens are currently being tested in the clinic against hematologic malignancies and solid tumors (5).
The ADCs currently undergoing clinical trials use various cytotoxic agents, which include microtubule-binding agents such as auristatins (MMAE and MMAF) and maytansinoids (DM1 and DM4), DNA topoisomerase-inhibiting agents such as SN38 and doxorubicin, and DNA-binding agents such as calicheamicin, duocarmycin, and pyrrolobenzodiazepine (1, 3, 6–10). The ADCs typically contain an average of about 4 molecules of a cytotoxic payload linked per antibody molecule, referred to as DAR, although some ADCs have a higher number of approximately 6 to 8 molecules linked per antibody (3, 4, 6, 7, 9). The payload molecules are attached at lysine or cysteine residues of the antibody via a linker moiety.
A number of noncleavable and cleavable linkers, with different mechanisms of degradation inside target cells, have been used in ADCs in the clinic. The noncleavable linkers—SMCC (used in ado-trastuzumab emtansine and other maytansinoid ADCs) and MC (employed in auristatin ADCs)—connect the payload to the antibody via a thioether bond. The noncleavable linker-bearing ADCs undergo proteolysis of their antibody component in lysosomes to release cytotoxic lysine- or cysteine-linked payload catabolites (11–13). Upon ultimate efflux out of target cells, these polar catabolite molecules, which bear both a positive and a negative charge, cannot easily enter and kill neighboring cells and therefore lack bystander-killing activity (11, 14, 15).
Cleavable linkers contain intralinker cleavage sites designed to allow efficient release of the payload in target cancer cells, while ideally being stable during circulation. Cleavable linker designs include disulfide, peptide, hydrazone, and carbonate linkages. The sterically hindered disulfide bond in the DM4 conjugate with SPDB or charged sulfo-SPDB linker is stable to cleavage during circulation, while allowing disulfide reduction and S-methylation of the polar lysine-linked DM4 catabolite inside target cells to generate uncharged DM4 and S-methyl DM4 molecules (11, 16, 17). Upon efflux from target cells, the uncharged maytansinoid molecules can enter and kill neighboring bystander cells that do not express target antigen (11, 14, 15). The VC-MMAE peptide linkage (used in brentuximab vedotin) is cleaved in lysosomes and upon further PABC spacer-immolation releases cytotoxic amine-bearing MMAE, which is cell-permeable and can exert bystander cytotoxic activity toward neighboring cells (12, 18). The calicheamicin ADCs use a hybrid linker containing both a stable disulfide and a hydrazone. The calicheamicin payload is released upon the cleavage of the hydrazone bond in the acidic pH of endosomes and lysosomes and further reduction of the disulfide bond (19, 20). A hydrazone linkage is used for anti-CD74-doxorubicin ADC (9). A carbonate linkage is used for anti-CEACAM5 and anti–TROP2-SN38 ADCs (21).
The selection of a linker for an ADC typically is based on the preclinical in vivo therapeutic index, which is defined as the ratio of the MTD to the minimum efficacious dose (MED). Although the clinical MTDs remain to be fully established for the cleavable and noncleavable linker-bearing ADCs across multiple targets, the preclinical MTDs are higher for ADCs with noncleavable linkers than for those with cleavable linkers (12, 22, 23). The in vivo antitumor activities of cleavable linker-bearing ADCs are often superior (lower MEDs) compared with those with noncleavable linkers, presumably aided by the bystander-killing activity of the cell-permeable payload molecules released from target cells (22, 24).
To explore peptide linkers in ADCs with the potential for efficient catabolism and thus, high antitumor activity, we designed a maytansinoid ADC with a new, triglycyl peptide linker (CX), which requires a single peptide bond cleavage in lysosomes to release the cytotoxic payload (Fig. 1). In contrast, noncleavable linker conjugates such as SMCC-DM1 or MC-MMAF require extensive proteolysis of folded protein backbone to release the payload. Furthermore, the lysosomal cleavage of the triglycyl peptide linker is expected to release a maytansinoid catabolite bearing a carboxylic acid group. A significant fraction of the carboxylic acid catabolite would be uncharged at the acidic lysosomal pH (∼4.5), and therefore able to diffuse out of the lysosomes more readily than the zwitterionic, charged catabolites generated from noncleavable linker ADCs. After export out of the lysosomes into cytosol, the maytansinoid carboxylic acid catabolite should inhibit microtubule dynamics and cause cell death. Upon final efflux from target cells, the carboxylic acid catabolite would be negatively charged at the extracellular pH of approximately 7.4 and therefore not expected to diffuse into and kill neighboring cancer cells (no bystander effect). Overall, an ADC utilizing the CX linker might be expected to generate both a higher amount of catabolite and a more efficient lysosomal efflux of catabolite into the cytoplasm in target cells compared with an ADC with a noncleavable linker such as SMCC-DM1 or MC-MMAF.
To explore these hypotheses, ADCs of anti-EGFR and anti-EpCAM antibodies were prepared with the maytansinoid payload DM1, using either CX (triglycyl) or a noncleavable SMCC linker. EGFR and EpCAM antigens are expressed on the cell-surface of multiple cancer cell lines and cancer tissues, making these suitable cancer targets for testing activities of ADCs (25–27). We compared the pre-clinical in vitro and in vivo activities, pharmacokinetics, and maximum tolerated doses of CX and SMCC linker-derived ADCs.
Materials and Methods
Human cancer cell lines—Calu-3, HCC827 (lung adenocarcinoma); PC-9, H1975 (non–small cell lung adenocarcinoma); H292 (mucoepidermoid pulmonary carcinoma); H226 (squamous cell lung carcinoma); SK-MES-1 (non–small cell lung squamous cell carcinoma); SW2 (small-cell lung adenocarcinoma); CAL 27, HSC-2, FaDu, SAS, Ca9-22, OSC19 (oral squamous cell carcinoma); COLO 205, HT-29, LoVo, HCT-15 (colorectal adenocarcinoma); A-431 (epidermoid carcinoma); RPMI-8226 (myeloma); OVCAR-3 (ovarian adenocarcinoma); BxPC3 (pancreas adenocarcinoma); and SK-MEL-28 (melanoma)—were obtained from the ATCC, JCRB, and Dr. E. Menta (LoVoDOX) within the period of 2000–2015. The cell lines were characterized by the vendor; no further cell line authentication was conducted. Upon receiving from the vendor, each cell line was expanded by passaging two to three times, aliquoted, and frozen. For use in experiments, cell lines were cultured in media recommended by the vendor in a humidified incubator at 37°C, 5% CO2 for no longer than 2 months. SMCC or sulfo-SMCC (4-(N-maleimidomethyl)cyclohexanecarboxylic acid N-hydroxysuccinimide ester, or the 3-sulfo-N-hydroxysuccinimide ester) were from Thermo Scientific. Chemical reagents were from Sigma-Aldrich. Humanized anti-EGFR antibody, humanized anti-CanAg antibody, chimeric mouse-human anti-EpCAM antibody, and chimeric nonbinding control antibody were all constructed with the human IgG1 isotype and generated at ImmunoGen, Inc. The syntheses of linkers and catabolites are described in Supplementary Method S1.
Preparation of conjugates
Conjugates of humanized or chimeric antibodies with maytansinoid (DM1) payloads were prepared using the heterobifunctional CX peptide linker bearing N-hydroxysuccinimide ester and maleimide groups, and SMCC (or sulfo-SMCC), similar to conjugation methods described previously (16, 17). The Ab-SMCC-DM1 and Ab-CX-DM1 conjugates were prepared with a typical average of about 3 to 4 DM1 molecules linked per antibody molecule (DAR). Ab-CX-DM1 conjugates were also prepared with a higher incorporation of 8 to 9 DAR, though not tested in vivo. All conjugates had high monomer composition (97%–99% monomer) and showed similar binding to antigen-expressing cell lines as unconjugated antibody.
Radiolabeled antibody-SMCC-[3H]DM1 and CX-[3H]DM1 conjugates were prepared by modifying the antibody with a 7- to 8-fold molar excess of a pre-mixed solution of SMCC-[3H]DM1 or CX-[3H]DM1 (prepared by reacting SMCC or CX linker with a 1.3-fold molar excess of [3H]DM1 (N2'-deacetyl-N2'-(3-mercapto-1-oxopropyl)-maytansine; bearing a tritium label at the C-20-methoxy group; ref. 14). The conjugation reaction was carried out at 2.5 mg/mL antibody concentration in N-(2-hydroxyethyl)piperazine-N'-(3-propanesulfonic acid; EPPS) aqueous buffer, pH 8, containing 5% to 10% N, N-dimethylacetamide (DMA). After overnight incubation at room temperature, conjugates were purified by gel filtration. The purified conjugates were analyzed for concentration of Ab and DM1 by UV/Vis spectrometry, percent monomeric conjugate by size-exclusion chromatography (SEC), and [3H] incorporation by liquid scintillation counting (LSC). The antibody-SMCC-[3H]DM1 and CX-[3H]DM1 conjugates contained 3 to 4 DAR and specific radioactivity of 0.7 Ci/mmol conjugated maytansinoid.
In vitro cytotoxicity of CX and SMCC ADCs
Cancer cells (about 3,000 cells/well) were incubated in a 96-well clear or white cell-culture plate (clear-bottom) with anti-EpCAM or anti-EGFR SMCC-DM1 and CX-DM1 conjugates at various concentrations (10−11–10−8 mol/L) in a 37°C incubator with 5% CO2. For anti-EpCAM conjugates, control wells contained a mixture of conjugate and 200-fold molar excess of unconjugated anti-EpCAM antibody. Nonbinding antibody conjugates were used as controls for anti-EGFR conjugates. After 4 to 5 days of continuous incubation with conjugates, WST-8 or CellTiter-Glo reagent (Dojindo or Promega) was added to wells, followed by the measurement of absorbance or luminescence. In separate experiments, a lysosomal inhibitor, Bafilomycin A1 (8 nmol/L), was investigated for its effect on the cell cycle arrest by ADCs (3–30 nmol/L), using a previously described method (11). The bystander cytotoxic activity of anti–EGFR-CX-DM1 was tested using a mixed culture of EGFR-positive and -negative cells as described in Supplementary Method S2.
Cellular binding studies
The binding of ADCs and labeled antibodies (phycoerythrin- or Alexa-labeled) to EGFR- or EpCAM-expressing cancer cells were evaluated using flow cytometry of trypsinized cells as described before (15). The EpCAM expression level of trypsinized Calu-3 cell line was measured as 3.6 × 105 antigens per cell. The reported EpCAM expression values for COLO 205, LoVoDOX, and RPMI-8226 cell lines are 9.3 × 105, 1.5 × 105, and 5 × 104 respectively (23). The EGFR expression levels were measured for several cell lines: HSC-2 (1 × 106), Ca9-22 (1 × 106), HCC827 (6.7 × 105), and H1975 (5 × 104).
Catabolism of CX and SMCC ADCs by cancer cells
The catabolites of [3H]-labeled and nonradiolabeled ADCs generated by cancer cells in 1 day were measured using HPLC or binding-competition ELISA methods described in the Supplementary Method S3 and as described previously (11, 28, 29).
In vivo activity of anti-EpCAM and anti-EGFR CX and SMCC ADCs
The antitumor activities of CX-DM1 and SMCC-DM1 conjugates derived from anti-EGFR, anti-EpCAM, or nonbinding antibody were evaluated in female nude mice bearing established xenograft tumors. For mice bearing Calu-3 xenografts, a single intravenous bolus administration of PBS, unconjugated anti-EpCAM antibody (15 mg/kg), anti–EpCAM-SMCC-DM1 (60, 200, and 300 μg of DM1/kg, approximately equivalent to 3, 10, and 15 mg/kg of protein), anti–EpCAM-CX-DM1 (60, 200, and 300 μg of DM1/kg), and nonbinding antibody-SMCC-DM1 or CX-DM1 conjugate (300 μg DM1/kg) were administered on day 5 after inoculation to 6 mice per treatment group. Tumor growth was monitored and tumor volume was calculated using the following formula: length × width × height × ½. Similar treatments were done for EGFR-expressing FaDu, HSC-2, and H1975 xenografts using anti-EGFR CX and SMCC ADCs.
Tolerability of CX and SMCC ADCs
The acute toxicity of CX-DM1 and SMCC-DM1 ADCs were evaluated in CD-1 mice based on monitoring body weight loss and clinical observations over 14 days following fractionated intravenous injections via tail vein of the ADCs at various doses up to a total dose of 3,000 μg/kg DM1 (∼150 mg/kg antibody). Mice were dosed 5 times at injections of up to 0.3 mL administered over 8 hours.
Pharmacokinetics of CX and SMCC ADCs
CD-1 mice were dosed with 10 mg/kg (i.v., single bolus, protein dose) of humanized anti-EGFR antibody-SMCC-[3H]DM1 and CX-[3H]DM1 conjugates (3.1–3.4 DAR). The antibody was developed against human EGFR and does not bind to mouse EGFR. Plasma samples were collected at various time intervals and conjugate concentrations were measured using LSC. Pharmacokinetic parameters were calculated using Pharsight WinNonLin software.
Preparation of maytansinoid ADCs with peptide (CX) and noncleavable (SMCC) linkers
New heterobifunctional linkers containing 2, 3, or 4 glycyl residues (1a–1c) or a valine–citrulline–glycine peptide, each bearing maleimide and N-hydroxysuccinimide ester, were synthesized (Fig. 1; Supplementary Method S1). ADCs derived from humanized or chimeric antibodies and maytansinoid payload were prepared using these CX peptide linkers or the noncleavable SMCC linker (Fig. 1). The conjugation strategy involved reaction of the thiol-bearing maytansinoid (DM1) with the maleimide group of the heterobifunctional CX or SMCC linker yielding a linker-DM1 adduct, which was conjugated with antibody at lysine residues to generate ADCs with an average load of 3-4 DAR. A 2-fold higher aqueous solubility was observed for the CX linker-DM1 adduct compared with the SMCC linker-DM1 adduct (0.2 and 0.1 mmol/L, respectively).
In vitro cytotoxic activities of CX and SMCC ADCs
In an initial experiment to determine the optimal number of glycyl residues for the peptide linkers, anti-EGFR antibody–maytansinoid conjugates with di-, tri-, and tetra-glycyl linkers were tested for in vitro cytotoxicity toward a panel of EGFR-expressing cancer cell lines (HSC-2, Ca9-22, PC-9, and H1975). The tri- and tetra-glycyl linked conjugates were found to have similar cytotoxic activities, but the di-glycyl linked conjugate was either similar or slightly worse in activity (Supplementary Fig. S1). Furthermore, the triglycyl linker conjugate showed overall similar cytotoxicity compared with a conjugate containing a valine–citrulline–glycine peptide linker in several cell lines tested (A-431, Ca9-22, HSC-2, H1975, and OSC19; Supplementary Fig. S1). On basis of these results, we selected the triglycyl linker as the lead peptide linker for all further studies.
We then compared the in vitro cytotoxic activities of anti-EGFR and anti-EpCAM antibody–maytansinoid conjugates prepared with the triglycyl linker (CX ADC) and with a traditional, noncleavable, SMCC linker (SMCC ADC) against a panel of cancer cell lines expressing EGFR and EpCAM, respectively. The unconjugated anti-EpCAM antibody, when added alone, lacked cytotoxic or inhibitory activity. The unconjugated anti-EGFR antibody, when tested alone at high concentrations (>1 nmol/L), partially inhibited the growth of some EGFR-expressing cell lines in vitro. The cytotoxic activities of the anti-EGFR ADCs, however, could be clearly distinguished from the inhibitory activity due to the anti-EGFR antibody component alone. Table 1 lists the IC50 values of anti-EGFR and anti-EpCAM ADCs with the CX or SMCC linker in several cell lines. Representative cytotoxicity profiles for a few cell lines at various concentrations of conjugates are shown in Fig. 2. A lysosomal mechanism of activation of CX and SMCC ADCs was supported by the blockade of their cell-cycle suppression by Bafilomycin A1, an inhibitor of lysosomal processing (Supplementary Method S2).
|.||.||IC50 (nmol/L) of ADC .||.|
|Antigen .||Cell line .||CX .||SMCC .||CX/SMCC activityb .|
|.||.||IC50 (nmol/L) of ADC .||.|
|Antigen .||Cell line .||CX .||SMCC .||CX/SMCC activityb .|
aCells were incubated with CX-DM1 and SMCC-DM1 ADCs continuously for 4–5 days. IC50 values are listed as “>” when IC50 could not be reached at the highest tested concentration.
bCX/SMCC activity ratio = IC50 SMCC/÷ IC50 CX. Activity is proportional to inverse of IC50.
cThe SMCC-linked EGFR ADC showed no cytotoxicity toward CAL 27 cells at the highest tested concentration (3 nmol/L) in a 4-day continuous incubation cytotoxicity assay. IC50 of CX ADC = 2 nmol/L.
Two patterns of cytotoxic activities were observed for CX- and SMCC-linked conjugates in the cell lines tested. The CX and SMCC ADCs had similar cytotoxic activities in one set of cell lines, whereas the CX ADC was more active than the SMCC ADC in another set of cell lines for both anti-EGFR and anti-EpCAM ADCs (Table 1; Fig. 2).
For anti-EGFR conjugates, the CX ADC showed similar cytotoxicity as the SMCC ADC in a number of EGFR-expressing cancer cell lines (H292, HCC827, H226, SK-MES-1, FaDu, BxPC3, and OSC19). In contrast, the CX ADC was more active than the SMCC ADC (8–100-fold lower IC50 for CX ADC) in several other cell lines, including HSC-2, A-431, Ca9-22, PC-9, SAS, and H1975 (Table 1, Fig. 2). The CX ADC of a nonbinding antibody was inactive even at a high concentration of 10 nmol/L (Supplementary Fig. S1).
The cytotoxic activities of anti-EpCAM SMCC and CX ADCs were similar in several cancer cell lines, including COLO 205, HT-29, LoVo, OVCAR-3, and HCT-15. However, the anti-EpCAM CX ADC was more active (5–13-fold lower IC50) than the SMCC ADC in another set of cancer cell lines, which included Calu-3, CAL 27, A-431, and RPMI-8226. In the multidrug-resistant cell line, LoVoDOX, which expresses high levels of P-glycoprotein (Pgp), the SMCC ADC was inactive at a high concentration of 5 nmol/L, whereas the CX ADC was active with an IC50 of 0.6 nmol/L (Table 1, Fig. 2). CX ADCs with high number of linked payload molecules (∼8 and 9.6 DAR) were extremely potent toward a relatively low antigen-expressing cell line, RPMI-8226, with an IC50 value of approximately 0.1 nmol/L, in comparison with the IC50 value of 0.3 nmol/L for an ADC with a typical DAR of approximately 3 to 4 (Supplementary Fig. S2). The cytotoxicities of all anti-EpCAM ADCs were antigen-specific and were abrogated when conjugates were mixed with excess, unconjugated anti-EpCAM antibody (Fig. 2).
Catabolism of SMCC and CX ADCs in target cancer cells
The catabolism of CX and SMCC ADCs by cancer cell lines was studied by three methods: (i) HPLC with radiometric detection (for [3H]-labeled ADCs), (ii) HPLC with UV detection (for nonradiolabeled ADCs), and (iii) binding-competition ELISA (for nonradiolabeled ADCs). The HPLC method with UV detection had low sensitivity and could be used only for one cell line with high antigen expression. The more sensitive methods of radiometric HPLC detection and binding-competition ELISA were employed for evaluating catabolites in multiple cell lines with a range of antigen expression using [3H]-labeled or nonradiolabeled ADCs, respectively (Fig. 3, Supplementary Table S1, Supplementary Method S3). The catabolite generation was studied by treatment of cells with saturating levels of ADCs for a short-term (20 minutes–2 hours), followed by washing to remove unbound conjugate and further incubation for 1 day to allow processing. The time period of 1 day for processing was selected as it allowed a sufficient level of conjugate catabolism without cell death (11, 28).
Processing of anti–EpCAM CX-[3H]DM1 by COLO 205 cells generated mainly a carboxylic acid catabolite (DM-CX2) and a smaller amount of a lysine-linked catabolite (DM-CX-lysine), arising from the cleavage of a peptide bond in triglycyl linker, or proteolysis of antibody at lysine site, respectively (Fig. 3A). Processing of CX ADC in Calu-3 cells generated two carboxylic acid catabolites (DM-CX1 and DM-CX2) arising from cleavage of triglycyl peptide linker at two different amide bonds (Fig. 3A). Lysine-linked CX catabolite was not observed in Calu-3 cells. The catabolism of anti-EpCAM SMCC-[3H]DM1 in both COLO 205 and Calu-3 cells generated only the lysine-linked catabolite (lysine-SMCC-DM1), consistent with the published catabolism data for SMCC ADCs (Fig. 3A; refs. 11, 28).
The catabolism of a nonradiolabeled anti-CanAg CX ADC in COLO 205 cells was analyzed using HPLC with UV detection. COLO 205 cells express high levels of CanAg antigen on their cell surface (∼4 × 106/cell) and generated sufficiently high levels of maytansinoid catabolites to allow HPLC with UV detection (15). The major catabolite was DM-CX2 carboxylic acid. The identity of the catabolite was confirmed by comparison of its mass spectrum and HPLC retention time with a synthesized standard (Fig. 3B).
A sensitive ELISA method based on maytansinoid-binding competition was used to quantitate the catabolites generated from CX and SMCC ADCs in several EGFR and EpCAM expressing cell lines (Supplementary Table S1). The ELISA inhibition curves were observed to be identical for the synthetic catabolite standards, DM-CX1, DM-CX2 (CX ADC catabolites), and lysine-SMCC-DM1 (SMCC ADC catabolite), thus allowing the use of a single catabolite standard to analyze multiple catabolites by ELISA (Supplementary Fig. S3).
The processing of anti-EGFR CX and SMCC ADCs by HSC-2 cells was analyzed using ELISA. The CX ADC was 13-fold more cytotoxic than the SMCC ADC in HSC-2 cells in a 5-day cytotoxicity assay (Table 1, Fig. 2A). Processing of the CX ADC in 1 day generated a total of 3.49 pmol catabolite per million cells, of which 0.72 pmol was in the cells and 2.77 pmol in the medium (Supplementary Table S1). In contrast, the SMCC ADC generated less total catabolite (2.03 pmol catabolite per million cells), of which a majority was in the cells (1.78 pmol) and very little in the medium (0.25 pmol), suggesting poor lysosomal export of the catabolite of SMCC ADC. In cell lines where the CX ADC was more cytotoxic than the SMCC ADC, the processing of CX ADC generated a greater amount of total catabolite and a greater proportion of the total catabolite was transported into the media than for SMCC ADC: HSC-2 and A-431 (EGFR) and Calu3 (EpCAM; Supplementary Table S1).
In contrast, in cell lines where both CX and SMCC ADCs were similarly cytotoxic, the processing of both CX and SMCC ADCs showed high levels of total catabolite and also significantly high levels of catabolite transport into the media: HCC-827 (EGFR) and COLO 205 (EpCAM; Supplementary Table S1). For example, in HCC827 cells where both CX and SMCC ADCs were similarly cytotoxic, the catabolite levels in cells were similar (1.06 and 1.10 pmol, respectively), and relatively high catabolite amounts were observed in the media for both CX and SMCC ADCs (1.66 and 0.88 pmol, respectively).
The bystander cytotoxicities of the major catabolites arising from CX and SMCC ADCs, DM-CX2 and lysine-SMCC-DM1, respectively, were compared with that of an uncharged maytansinoid molecule, S-methyl DM4, a catabolite of disulfide-linked maytansinoid ADCs. Upon addition of the chemically synthesized catabolites to the medium in several cell lines (HSC-2, SK-MES-1, PC-9, and SK-MEL-28), the IC50 values of the charged DM-CX2 and lysine-SMCC-DM1 catabolites (28–120, and 56–290 nmol/L, respectively) were >100-fold less potent compared with that of the uncharged S-methyl DM4 (IC50 = 0.07-0.27 nmol/L; data not shown; ref. 30). Furthermore, the anti-EGFR CX ADC did not show bystander cytotoxic activity toward EGFR-negative Ramos cells in a mixed cell culture containing EGFR-positive Ca9-22 cells, in contrast with a disulfide-linked conjugate that showed bystander cytotoxic activity (Supplementary Fig. S4). The catabolites of CX and SMCC ADCs upon efflux from target cells would be charged at the extracellular pH of 7.4, and therefore not expected to be cell-permeable and cytotoxic toward bystander cells.
In vivo antitumor activities of anti-EpCAM and anti-EGFR ADCs with SMCC or CX linker
The antitumor activities of anti-EpCAM and anti-EGFR ADCs with CX or SMCC linker (CX-DM1 or SMCC-DM1) were evaluated in immunocompromised mice bearing established xenografts (6 mice/treatment group). Anti-EpCAM ADCs were tested in mice bearing established Calu-3 human lung adenocarcinoma xenografts (Fig. 4A). Treatments with a single dose of unconjugated anti-EpCAM antibody or the nonbinding antibody-SMCC-DM1 and CX-DM1 ADCs at 15 mg/kg (Ab dose) were ineffective in this study, showing test/control tumor ratio (T/C) of >42%. The anti–EpCAM-SMCC-DM1 was active at the highest administered dose of 15 mg/kg with a T/C of 39%, although it did not result in any partial (PR) or complete regressions (CR). In contrast, treatment of mice with anti–EpCAM-CX-DM1 was highly effective at killing Calu-3 tumors. At the lowest dose tested, 3 mg/kg, the T/C was 10%, with 1 of 6 PR. Higher doses (10 or 15 mg/kg) were even more efficacious, resulting in T/C values of 0 and 1%, respectively. In addition, there were PR (5/6, 6/6) and CR (4/6, 4/6) observed at the 10 or 15 mg/kg dose, respectively.
The anti-EGFR CX and SMCC ADCs were tested in FaDu (oral squamous cell carcinoma), HSC-2 (oral squamous cell carcinoma), and H1975 (non–small cell lung adenocarcinoma) xenograft models (Fig. 4B–D). Unconjugated anti-EGFR antibody was also tested at similar doses.
In the FaDu model, mice were dosed with unconjugated anti-EGFR antibody, anti–EGFR SMCC-DM1 or CX-DM1 conjugate at 5 mg/kg. The unconjugated anti-EGFR antibody was active with a T/C of 35%, with no PR or CR. Conjugation of DM1 via either SMCC or CX linker enhanced the activity of the anti-EGFR antibody, with T/C values of 19% for both ADCs. At the tested dose, there were no PR or CR in the SMCC-DM1 treatment group and one PR in the CX-DM1 treatment group (Fig. 4B).
In the HSC-2 model, mice were treated with anti-EGFR, anti–EGFR SMCC-DM1 or CX-DM1 conjugate at 5 mg/kg. The unconjugated anti-EGFR antibody was active, with a T/C of 39%; however, no PR or CR was observed. The SMCC-DM1 and CX-DM1 conjugates were both highly active in this model with T/C values of 6% and 0%, respectively. Treatment with the CX-DM1 ADC resulted in many more responses than the SMCC-DM1 treatment (6/6 vs. 2/6 PR, and 6/6 vs. 1/6 CR, respectively). In addition, progression-free state lasted significantly longer in the CX-DM1–treated mice (Fig. 4C).
Finally, in the H1975 model, mice were treated with unconjugated anti-EGFR antibody, anti-EGFR SMCC-DM1 or CX-DM1 conjugate at single dose of 10 mg/kg. The unconjugated antibody was inactive in this model with a T/C of 75%. Treatment with anti–EGFR-SMCC-DM1 conjugate was minimally active (T/C of 41%, 1/6 PR, 1/6 CR). In contrast, the anti–EGFR-CX-DM1 conjugate was highly active, with a T/C of 12%, and with 6/6 PR and 3/6 CR (Fig. 4D). Similar to the HSC-2 study, treatment with the CX-DM1 ADC resulted in a longer period of tumor regression compared to the SMCC-DM1 ADC (Fig. 4D).
MTD of CX and SMCC ADCs in mice
The acute tolerability of CX and SMCC ADCs was evaluated in female CD-1 mice, assessed by measurements of body weight change and clinical observations (8 mice/treatment group). The SMCC ADC dosed at 3 mg DM1/kg (∼150 mg/kg Ab) resulted in an average nadir body weight loss of 10.8% (range, 7.7%–16%) with 0/8 animal deaths. At the same dose, treatment with the CX ADC resulted in a similar nadir body weight loss of 9.9% (7.5%–17%) with 0/8 animal deaths. The nadir for body weight loss for both ADCs occurred between days 3 to 5 (Fig. 5A). The MTD of CX ADC (about 150 mg/kg Ab) was about 50-fold higher than the minimum efficacious dose of CX ADC (3 mg/kg Ab) in the tested xenograft models (Fig. 4), affording a high preclinical therapeutic index.
Pharmacokinetics of CX and SMCC ADCs
Radiolabeled anti-EGFR antibody-SMCC-[3H]DM1 and CX-[3H]DM1 conjugates were administered to CD-1 mice as a single bolus injection of 10 mg/kg (protein dose) and plasma samples collected at several time points. Conjugate concentrations in plasma samples were measured by 3H-scintillation counting and pharmacokinetic parameters were calculated (Fig. 5B).
The clearance curves for the two ADCs were characterized by biphasic pharmacokinetics, with an initial rapid decline in concentration (distribution phase), followed after about 24 hours by a slower rate of decline (terminal elimination phase), which followed first-order kinetics. The pharmacokinetic parameters for the two ADCs were comparable. For Ab-SMCC-DM1, the t½, CL, and AUC0-∞ were estimated to be 10.4 days, 0.7 mL/h/kg, and 15,225 h·μg/mL, respectively. The distribution volume at steady-state (VSS) was estimated to be 230 mL/kg, whereas the Cmax was 129 μg/mL (measured at the 2 minutes time point). For the Ab-CX–DM1 conjugate, the t½, CL, and AUC0-∞ were estimated to be 9.9 days, 0.7 mL/h/kg, and 14,370 h·μg/mL, respectively. The VSS was estimated to be 232 mL/kg, whereas the Cmax was 124 μg/mL (measured at the 2 minutes time point). Both Ab-SMCC-DM1 and Ab-CX-DM1 were stable in circulation in mice, showing PK parameters similar to those previously reported for trastuzumab–SMCC-DM1 conjugate (28, 31).
A new, triglycyl peptide linker (CX) was used in ADCs to improve both lysosomal proteolytic release and transport of cytotoxic payload catabolites into the cytoplasm. The activity and processing of ADCs prepared with CX linker were compared with those bearing a noncleavable SMCC linker. Incorporation of the CX linker in anti-EGFR and anti-EpCAM ADCs led to improvement in cytotoxic potencies in a number of cell lines in vitro compared with the respective ADCs bearing the noncleavable SMCC linker. Furthermore, in several EGFR-expressing xenograft models and in an EpCAM-expressing xenograft model, the CX ADC was significantly more active in vivo than the SMCC ADC.
The SMCC ADC requires lysosomal proteolysis of the antibody scaffold to release a polar catabolite, lysine-SMCC-DM1, which bears both a negative and a positive charge (11, 14, 28). In contrast, the lysosomal processing of the CX ADC requires cleavage at only one of the amide bonds in the triglycyl linker to generate maytansinoid catabolites bearing carboxylic acid group (DM-CX1, DM-CX2). The triglycyl linker conjugate showed similar cytotoxicity as valine–citrulline–glycine linker conjugate in a number of cell lines tested, indicating an efficient cleavage of the triglycyl linker in lysosomes without specifically requiring Cathepsin B. Target cancer cell lines with deficient or limited proteolytic processing could generate higher levels of catabolites from CX ADC than from SMCC ADC, resulting in a greater activity for CX ADC than for SMCC ADC.
Lysosomes are acidic vesicles, with a pH value of about 4.5 (32). The DM-CX1 and DM-CX2 catabolites of CX ADCs contain carboxylic acid groups with estimated pKa values of about 4.8 and 3.4, respectively (33, 34). A significant proportion of the CX ADC catabolites, therefore, would be uncharged in the acidic lysosomes. The rate of efflux of the uncharged CX ADC catabolites out of the lysosomes is expected to be greater than that for the charged lysine-SMCC-DM1 catabolite, which could be transported slowly out of lysosomes in some cells. In a recent report, knockdown of a lysosomal membrane protein (SLC46A3) resulted in the abrogation of cytotoxicity of SMCC-DM1 conjugates in several cell lines (35). Cell lines with deficient lysosomal transport of lysine-SMCC-DM1 catabolite (due to low levels of lysosomal export proteins such as SLC46A3) could possibly be sensitive to killing by CX ADC as the uncharged DM-CX1 and DM-CX2 catabolites may diffuse out of acidic lysosomes without requiring specific lysosomal transporters.
The in vitro cytotoxicity is typically measured after 4 to 5 days of continuous incubation of cells with ADCs, allowing sufficient time for killing of cells during cell division. The amounts of catabolites in cells and in media are measured after a 1-day period of processing, at which time the cells are still intact, allowing a measurement of the transport of catabolites in live cells. On the basis of previous studies of catabolism of maytansinoid and MMAE conjugates in non–multidrug-resistant cell lines with high levels of antigen, total catabolite amounts generated were greater than 1 pmol per million cells (10−18 mol/cell), with about 0.2 to 0.3 μmol/L intracellular catabolite concentration and also significant levels of catabolites observed in the medium (18, 29). The study uses the measurement of catabolites in medium as a convenient, indirect method of measuring export of catabolites out of the lysosomes into the cytoplasm and subsequent efflux out of the cell into the medium.
The cytotoxic potency of the CX and SMCC ADCs correlated with the catabolite level and efflux. In cell lines that showed greater cytotoxicity for CX ADC compared with SMCC ADC, both a greater level of total catabolite and a higher proportion of catabolite in media were observed for CX ADC. The catabolite of SMCC ADC could be trapped in the lysosomes in these cells, thus not being exported into the cytoplasm (the site of microtubule inhibition) or further into the medium. In other cell lines where both CX and SMCC ADCs were similarly cytotoxic, the levels of catabolites in cells were similar and significant amounts were observed in the media for both CX and SMCC ADCs, suggesting efficient catabolite transport for both ADCs in these cells. The CX ADC, similar to SMCC ADC, does not have bystander-killing activity toward neighboring tumor cells. Upon ultimate release from the target cell, the CX ADC catabolite would be present mostly in the charged carboxylate form at the extracellular pH of approximately 7.4 and not readily diffuse into neighboring cells. In contrast, maytansinoid ADCs with alternate linkers such as disulfide or anilino peptide generated nonpolar catabolites with bystander-killing activity (22, 36).
The charged, CX and SMCC unconjugated catabolites are cytotoxic upon extracellular addition into the medium at relatively high concentrations (IC50 ≈ 0.03–0.12 μmol/L and 0.06–0.3 μmol/L, respectively). Similar high concentrations of catabolites are generated within the target cells from ADC processing (about 0.2 μmol/L intracellular catabolite concentration), with significant catabolite export into the medium (29). This suggests a similar mechanism of transport of charged catabolites out of the cytoplasm into the medium or from the medium into the cytoplasm at these high concentrations. The charged catabolites of CX and SMCC ADCs are diluted upon release into the extracellular space, and therefore cannot enter and kill neighboring cells.
The CX ADC showed greater activity and higher therapeutic index over SMCC ADC in several preclinical xenograft models for two different antigens, EGFR and EpCAM. The CX ADC showed either greater activity than the SMCC ADC at a lower dose, or resulted in a longer progression-free state or greater PR and CR at a similar dose. In an EpCAM-expressing xenograft model, a high preclinical therapeutic index (MTD/MED ratio) of 50-fold was observed for CX ADC, which was even more active at a low dose of 3 mg/kg compared with the 15 mg/kg dose of SMCC ADC. Both SMCC and CX ADCs showed stable pharmacokinetics in mice, demonstrating that the CX linker was stable in plasma during circulation.
The CX linker is a promising new linker for ADCs. The CX ADC offers the potential of improved activity compared with the ADC using the noncleavable SMCC linker, while maintaining a similar preclinical MTD as SMCC ADC. ADCs with CX linker could potentially be used for the treatment of tumors with high, homogeneous levels of antigen and for hematologic malignancies, where bystander-killing activities of ADCs may not be required.
Disclosure of Potential Conflicts of Interest
J.M. Lambert has ownership interest (including patents) in ImmunoGen, Inc. No potential conflicts of interest were disclosed by the other authors.
Conception and design: R. Singh, Y.Y. Setiady, J. Ponte, W.C. Widdison
Development of methodology: R. Singh, E.E. Hong, L. Dong, K. Veale, J.A. Costoplus, W.C. Widdison
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Singh, J. Ponte, K.C. Lai, E.E. Hong, L. Dong, G.E. Jones, J.A. Coccia, K. Veale, A. Skaletskaya, R. Gabriel, P. Salomon, R. Wu, Q. Qiu
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Singh, Y.Y. Setiady, E.E. Hong, L. Dong, G.E. Jones, J.A. Coccia, L. Lanieri, P. Salomon
Writing, review, and/or revision of the manuscript: R. Singh, Y.Y. Setiady, J. Ponte, Y.V. Kovtun, K.C. Lai, E.E. Hong, N. Fishkin, L. Dong, P. Salomon, J.M. Lambert, R.V.J. Chari
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Singh, Y.Y. Setiady, L. Lanieri, Q. Qiu
Study supervision: R. Singh, Y.Y. Setiady, J. Ponte, E.E. Hong, L. Lanieri, H.K. Erickson, J.M. Lambert, R.V.J. Chari
The authors thank Thomas Chittenden, Thomas Keating, and Richard Bates for their careful review of the article.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.